基于贝叶斯法则的空间自相关误差自相关模型变量选择研究

Variable Selection of Spatial Autoregressive Model with Autoregressive Disturbances by Bayes Rule

本文将研究贝叶斯法则视角下的空间自相关误差自相关模型(Spatial Autoregressive Model with Autoregressive Disturbances,SARAR模型)变量选择问题。通过将基于BIC准则的子集选择法推广到空间模型,实现SARAR模型的变量选择,并证明在一定条件下,对于SARAR模型的变量选择BIC准则具有良好的渐近性质。同时本文还将利用Monte Carlo模拟验证BIC准则能够很好的实现SARAR模型的变量选择。最后以股票收益率为例,在验证股票收益率具有空间效应的前提下,利用BIC准则对影响股票收益率的众多财务指标进行变量选择。

This paper studies the variable selection problem of spatial autoregressive model with autoregressive disturbances under the perspective of Bayes rule. We will apply the subset selection method with BIC criterion of linear model to spatial model, and prove that BIC rule has well asymptotic properties in spatial models under certain conditions. At the same time we will use Monte Carlo simulation to show that BIC criterion will be effective in variable selection of the model. At the end of the paper, we will use stock yield as the example. We will demonstrate the stock yield has the spatial properties, and then use the above method to select the factors of stock yield.